Toward scalable universal solvers for linear systems

Abstract

This proposal will investigate new algorithms for solving large-scale linear systems that arise in physical simulation and in data analysis. The goal is to develop a scalable solver that can address a wide range of fundamental problems. In particular, we will design practical, provable algorithms for solving connection Laplacian and graph Laplacian systems. We will implement these algorithms and demonstrate improved performance for applications in fluid dynamics and graphics processing. The technical approach is based on operator-adapted wavelets, called gamblets. We will use randomization to extend the gamblet decomposition to a wider class of operators

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 10, 2018
Source ID
N000141812363

Entities

People

  • Joel Tropp

Organizations

  • California Institute of Technology
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Linear Algebra